- Updated: April 6, 2026
- 2 min read
RightNow AI Unveils AutoKernel: Open‑Source Framework for Autonomous GPU Kernel Optimization
RightNow AI Unveils AutoKernel: Open‑Source Framework for Autonomous GPU Kernel Optimization
RightNow AI has announced the release of AutoKernel, an open‑source framework that brings an autonomous agent loop to GPU kernel optimization for any PyTorch model. The project, made publicly available on GitHub, aims to simplify the traditionally complex and manual process of tuning GPU kernels, allowing developers to achieve higher performance without deep hardware expertise.
AutoKernel works by automatically generating, profiling, and selecting the most efficient kernel implementations for a given model. It leverages a search‑and‑optimize loop that iteratively refines kernel parameters, ensuring optimal utilization of GPU resources across a wide range of hardware configurations.
Key features of AutoKernel include:
- Model‑agnostic optimization: Supports arbitrary PyTorch models, from simple CNNs to large transformer architectures.
- Zero‑code integration: Developers can plug AutoKernel into existing training pipelines with a single API call.
- Cross‑GPU compatibility: Works on NVIDIA, AMD, and Intel GPUs, automatically adapting to each platform’s quirks.
- Open‑source licensing: Released under the Apache 2.0 license, encouraging community contributions and extensions.
The framework is positioned as a response to the growing demand for automated performance tuning in AI workloads, where manual kernel engineering can become a bottleneck. By automating this step, RightNow AI hopes to accelerate model deployment and reduce the time‑to‑market for AI‑driven products.
For developers interested in trying out AutoKernel, the source code, documentation, and example notebooks are available on the official GitHub repository. The project also includes a set of benchmark results that demonstrate up to a 30% speed‑up on common vision and language models compared to baseline PyTorch kernels.
Read the original announcement on MarkTechPost for more technical details.
Explore related content on our site:
Stay tuned for future updates as the community begins to contribute enhancements and new features to AutoKernel.